Research On Control Technology Of Rehabilitation Gloves Based On Surface EMG Signals

Stroke is currently the leading disability factor for adults in the world,and survivors still have a 50%-70% probability of hemiplegia.Pneumatic rehabilitation gloves are a better solution for hand function rehabilitation,which is the most difficult postoperative rehabilitation for stroke patients.Different from the current pre-programming mode and the master-slave control mode,the myoelectric control mode has the advantages of good bionics,simple device structure,flexible application,etc.This subject aims to combine pneumatic transmission with myoelectric control technology and realize myoelectric control of pneumatic rehabilitation gloves.According to the characteristics of surface EMG signals,a dual-channel s EMG conditioner was designed.The surface EMG signals of finger flexors and extensors under different gestures were collected.A sliding window method for signal processing was used to split signals.A double threshold method was designed to detect the active segments.Besides,time-domain characteristics and decomposition coefficient characteristics of the wavelet packet of dual-channel surface EMG signals were extracted.A BP neural network classifier was designed and trained and the classifier was applied to perform offline recognition experiments on test samples.Furthermore,the adaptability of the pattern recognition algorithm was tested on ten experimenters.Servo control of the bending angle of a pneumatic rehabilitation glove was researched.A closed-loop control system for bending angle was designed based on fuzzy-PID switching control strategy and an experimental platform was built.Separate bending angle step and ramp signals following experiments were carried out.Based on the above researches,the threshold control strategy and the pattern recognition control strategy of myoelectric control methods were designed.Firstly,the online recognition algorithm of hand flexion and extension intention was designed in MATLAB/Simulink based on a threshold control strategy.And the slow flexion and extension of a rehabilitation glove were realized combined with servo control of the bending angle of the pneumatic soft actuators.Secondly,the online pattern recognition algorithm of gestures was designed in MATLAB/Simulink based on the pattern recognition strategy.And the multi-mode myoelectric control of gestures was realized.In the experiment,the online recognition includes five gestures:‘index finger pinching’,‘middle finger pinching’,‘palm bending’,‘palm extending’,and‘palm rest’.Finally,the corresponding solenoid valves were controlled according to the online recognition results,and a pneumatic rehabilitation glove was driven to perform the corresponding action mode.The effect of the online recognition algorithm on gestures of different experimenters was tested,the average online recognition rate was 75.45%.